Experts have mixed opinions on the extent to which generative AI will enhance productivity and the value it can provide relative to its costs, as it faces further challenges from chip and power shortages. 

In a report last week, Goldman Sachs looked into the implications of generative AI and asked industry and economic experts if the large degree of AI spending is likely to pay off. The promise of AI continues to be highlighted as large tech companies and other corporations will spend an estimated US$1 trillion on capital expenditures (CapEx) in the coming years, the report said, which includes investments in data centers, chips and other elements of AI infrastructure. 

“But this spending has little to show for it so far beyond reports of efficiency gains among developers. And even the stock of the company reaping the most benefits to date— Nvidia—has sharply corrected,” the report said. 

The report highlights an interview with Daron Acemoglu, a professor at MIT, who says he is skeptical about the transformational powers of AI. 

“He estimates that only a quarter of AI-exposed tasks will be cost-effective to automate within the next 10 years, implying that AI will impact less than five per cent of all tasks,” the report said. 

Acemoglu also has questions regarding whether AI will “create new tasks and products.” 

“So, he forecasts AI will increase U.S. productivity by only 0.5 per cent and GDP (gross domestic product) growth by only 0.9 per cent cumulatively over the next decade,” the report said. 

Jim Covello, the Goldman Sachs head of global equity research, said in the report that AI “must be able to solve complex problems” in order to “earn an adequate return” on the estimated $1 trillion price tag of developing and operating the technology. 

However, Covello says he doesn’t believe AI is built to solve complex problems. 

“He’s also doubtful that AI will boost the valuation of companies that use the tech, as any efficiency gains would likely be competed away, and the path to actually boosting revenues is unclear, in his view,” the report said. 

Despite questions among experts that AI will meet lofty expectations, some are optimistic about the technology. Joseph Briggs, a senior global economist at Goldman Sachs, shared a more favourable view of AI in the report. 

“He estimates that gen AI will ultimately automate 25 per cent of all work tasks and raise U.S. productivity by nine per cent and GDP growth by 6.1 per cent cumulatively over the next decade,” the report said. 

Briggs notes that although automating many “AI-exposed tasks” is not cost-effective today, he points to the likelihood that costs will drop over the longer term, allowing for more AI automation. 

The report also states Eric Sheridan, a U.S. internet equity research analyst at Goldman Sachs, is still enthusiastic regarding AI‘s transformative potential, noting that investors are “rewarding only those companies that can tie a dollar of AI spending back to revenues.” 

Power and chip shortages? 

Also explored in the report, was the possibility that even if AI has the ability to “generate significant benefits for economies and returns for companies,” shortages of key inputs like chips and power could prevent the technology from reaching its potential. 

The report cites the opinions of Goldman Sachs U.S. semiconductor analysts Toshiya Hari, Anmol Makkar and David Balaban who state that the number of chips is likely to “constrain AI growth over the next few years,” as demand surpasses supply. 

“But the bigger question seems to be whether power supply can keep up,” the report said. 

“GS (Goldman Sachs) U.S. and European utilities analysts Carly Davenport and Alberto Gandolfi, respectively, expect the proliferation of AI technology, and the data centers necessary to feed it, to drive an increase in power demand the likes of which hasn’t been seen in a generation.” 

 Rising power demands from AI pose a challenge for the U.S. power grid, according to Brian Janous, the co-founder of Cloverleaf Infrastructure and former VP of Energy at Microsoft. 

The report notes Janous’ position that U.S. utilities are not prepared for the “coming demand surge” and the aging U.S. power grid has not seen a rise in electricity consumption in nearly two decades.   

As a result of those constraints, Janous and Davenport say that “substantial investment” in U.S. power infrastructure will take a significant amount of time given the “highly regulated nature of the utilities industry and supply constraints.” The report says Janous warns that the “power crunch” could weigh on AI growth going forward.